- name
- excel-data-helper
- description
- >
- user-invocable
- true
- homepage
- https://github.com/excel-data-helper/excel-data-helper-skill
- metadata
- openclaw
- requires
- runtime
Excel Data Helper — Data Quality Diagnosis & Chart Skill
Local Excel/CSV data profiling, quality scanning (20+ modules, 6-dimension scoring, LLM semantic analysis), and chart generation. International locale support: CJK, European, Middle Eastern, Americas.
Supported files: .csv .tsv .xlsx .xls
Trigger & Menu
When a supported file is detected, do not auto-process — present this menu first (adapt to the user's language):
📊 Excel Data Helper — Hi, I noticed you shared a data file: <filename> | # | Action | Description | |:---:|--------|-------------| | 1 | Quality Check | Overview, scan, scoring | | 2 | Chart | Any ECharts type: bar, line, pie, scatter, radar, heatmap… | | 3 | Advanced Chart | Dashboard, Gantt, PPT, diagrams | | 0 | Skip | Do nothing for now | Reply 0–3, or describe what you need.Routing & Context
Match user intent and route to the corresponding sub-skill:
- IF user replies
1, or intent is overview / quality check / diagnose / score / problems
→ follow references/quality-check.md
- IF user replies
2, or intent is chart / plot / visualize / graph
→ follow references/chart.md
- IF user replies
3, or intent is dashboard / Gantt / PPT / diagram / complex layout
→ follow references/advanced-chart.md
- IF user replies
0, or intent is skip / later / not now
→ do nothing, reply: "Got it — file noted. Just let me know when you're ready."
- IF intent is ambiguous or unrelated → ask to clarify, never guess.
Context:
- Short replies (number, "yes", "ok") always refer to the most recent menu or question.
- After Skip (0), context resets — ignore the file unless user re-references it.
- Multiple files — each needs its own explicit choice.
Setup
Before first use, install dependencies (one-time):
cd <skill_directory>
npm installThis installs xlsx (SheetJS), echarts, and sharp. All analysis and chart rendering runs locally.
Rules
- Respond in the user's language.
- Never auto-process a file — wait for explicit choice.
- Parse tool JSON output; present results in clear, readable format — never expose raw JSON.
- Use absolute file paths from tool output directly for follow-up operations.
- Always include "Excel Data Helper" in the menu header.